|
|
 |
| |
|
|
|
An Improved Volatility Forecasting with GMLE-Based MS-GARCH Models: Evidence from BWP/USD, ZAR/USD, and BTC/USD |
|
|
|
PP: 343-360 |
|
|
|
doi:10.18576/jsap/140602
|
|
|
|
Author(s) |
|
|
|
Edwin Moyo,
Katleho Makatjane,
Ramasamy Sivasamy,
|
|
|
|
Abstract |
|
|
| In this study, the growing interconnectedness of the world’s financial markets is recognised, and we seek to investigate the complicated volatility dynamics and downside risk inherent in BWP/USD, ZAR/USD, and BTC/USD exchange rate returns. Aiming to overcome the shortcomings of conventional methods, this work seeks to provide a strong and computationally efficient approach for predicting and modelling volatility, and this approach is two-stage. We first use the GMLE method to estimate parameters within MS-GARCH models. The second part of the research evaluates how well the Student’s t, skewed Student’s t, and generalised error distribution capture the stylised features of financial returns. To reduce regime path dependency, the Hamilton filter is used; quasi- likelihood ratio tests verify statistically meaningful regime switching. Using AIC and BIC for model selection, the MS-GJR-GARCH model is found to be the best fit for BTC/USD, featuring two regime parameters and a skewed Student t distribution. By contrast, the MS-EGARCH specification is recommended for BWP/USD and ZAR/USD with the same student t distribution. With BTC/USD showing far more negative risk, especially over long forecasting horizons, empirical results show different volatility patterns across all three exchange rates. These findings draw attention to the increased investment risk of Bitcoin. The study finds that a strong foundation for modelling exchange rate volatility is provided by regime-switching models, including generalised maximum likelihood estimation and flexible error distributions. Future studies should look at macroeconomic factors that influence regime changes and expand the investigation to include additional developing market currencies. |
|
|
|
|
|
|
|
